11from pathlib import Path
12from typing import Dict, Union, Tuple, Optional
13
14import torch
15from torch import nn
16
17from labml import monit, lab, logger
18from labml.logger import Text, inspect
19from labml.utils.download import download_fileParent url
22CHECKPOINTS_URL = 'https://mystic.the-eye.eu/public/AI/models/GPT-NeoX-20B/slim_weights/'
23
24_CHECKPOINTS_DOWNLOAD_PATH: Optional[Path] = NoneDownload path
28def get_checkpoints_download_path():
29    global _CHECKPOINTS_DOWNLOAD_PATH
30
31    if _CHECKPOINTS_DOWNLOAD_PATH is not None:
32        return _CHECKPOINTS_DOWNLOAD_PATH
33
34    _CHECKPOINTS_DOWNLOAD_PATH = lab.get_data_path() / 'neox_fast' / 'slim_weights'
35    if not _CHECKPOINTS_DOWNLOAD_PATH.exists():
36        _CHECKPOINTS_DOWNLOAD_PATH = lab.get_data_path() / 'neox' / 'slim_weights'
37    inspect(neox_checkpoint_path=_CHECKPOINTS_DOWNLOAD_PATH)
38
39    return _CHECKPOINTS_DOWNLOAD_PATH42def get_files_to_download(n_layers: int = 44):48    layers = (Embedding layer
50            [0] +Transformer layers
52            list(range(2, 2 + n_layers)) +Final normalization layer and readout layer
54            [47, 48]
55    )
56
57    return (Vocabulary and configs
59            ['20B_tokenizer.json', 'configs/20B.yml', 'latest'] +Layer checkpoints
61            [f'global_step150000/layer_{i :02d}-model_{p :02d}-model_states.pt' for i in layers for p in range(2)] +Empty states (not used)
63            [f'global_step150000/mp_rank_{i :02d}_model_states.pt' for i in range(8)]
64    )67def download(n_layers: int = 44):Get files to download
73    files = get_files_to_download(n_layers)Iterate
76    for i, f in monit.enum('Download All', files):Log
78        logger.log(['Downloading ', (f'{i + 1 :3d}/{len(files)}', Text.meta), ': ', (f, Text.value)])Download
80        download_file(CHECKPOINTS_URL + f, get_checkpoints_download_path() / f)83def load_checkpoint_files(files: Tuple[str, str]):90    checkpoint_path = get_checkpoints_download_path() / 'global_step150000'
91    with monit.section('Load checkpoint'):
92        data = [torch.load(checkpoint_path / f) for f in files]
93
94    return dataparam
  is the parameter key
  is the name of the parameter p1
  first partition dictionary p2
  second partition dictionary97def merge_params_dim_0(param: Union[nn.Parameter, torch.Tensor], key: str, p1: Dict[str, torch.Tensor],
98                       p2: Dict[str, torch.Tensor]):107    w1, w2 = p1[key], p2[key]
108    param.data[:w1.shape[0]] = w1
109    param.data[w1.shape[0]:] = w2param
  is the parameter key
  is the name of the parameter p1
  first partition dictionary p2
  second partition dictionary112def merge_params_dim_1(param: Union[nn.Parameter, torch.Tensor], key: str, p1: Dict[str, torch.Tensor],
113                       p2: Dict[str, torch.Tensor]):122    w1, w2 = p1[key], p2[key]
123    param.data[:, :w1.shape[1]] = w1
124    param.data[:, w1.shape[1]:] = w2This does a sanity check to make use both partitions are the same
param
  is the parameter key
  is the name of the parameter p1
  first partition dictionary p2
  second partition dictionary127def merge_params_duplicate(param: Union[nn.Parameter, torch.Tensor], key: str, p1: Dict[str, torch.Tensor],
128                           p2: Dict[str, torch.Tensor]):139    w1, w2 = p1[key], p2[key]
140
141    diff = sum((w1 - w2) ** 2).item()
142    assert diff < 1e-4, f'The partitions do not match: {key}'
143
144    param.data[:] = (w1 + w2) / 2.param
  is the parameter key
  is the name of the parameter p1
  first partition dictionary p2
  second partition dictionary147def merge_params_sum(param: Union[nn.Parameter, torch.Tensor], key: str, p1: Dict[str, torch.Tensor],
148                     p2: Dict[str, torch.Tensor]):157    w1, w2 = p1[key], p2[key]
158
159    param.data[:] = w1 + w2163if __name__ == '__main__':
164    download()